# Research Papers — 2026-05 · bundle page 261 of 326 > Full-text markdown of papers #13001–13050 of 16266 published in 2026-05, newest first. AI-readable mirrors of the full 16266-paper 2026-05 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-05/page/261 [← Previous bundle](https://aipod.dev/papers/archive/2026-05/page/260.md) [Next bundle →](https://aipod.dev/papers/archive/2026-05/page/262.md) ### FlightSense: An End-to-End MLOps Platform for Real-Time Flight Delay Prediction via Rotation-Chain Propagation Features and Agentic Conversational AI Authors: Aditi J. Shelke, Renuka J. Shelke, Yash M. Kamerkar, Nitin P. Hazarani Published: 2026-05-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.07364v2 Flight delays impose cascading operational and financial burdens across the aviation network, costing the U.S. economy billions of dollars annually by disrupting interconnected aircraft rotation systems. While prior machine learning approaches have demonstrated strong predictive performance, most treat upstream delays as static input variables rather than explicitly modeling how delays propagate dynamically through aircraft rotation chains, and none have deployed such systems alongside a live weather-aware conversational AI interface for end-user interaction. This paper presents FlightSense, an end-to-end MLOps platform for real-time flight delay prediction built through a progressive three-version feature engineering framework. Version 1 trains an XGBoost classifier on 11 schedule-based features establishing a baseline ROC AUC of 0.732 on 7.07 million BTS 2018 On-Time Performance records. Version 2 introduces 11 delay propagation features derived from aircraft rotation chains via tail-number tracking, yielding the dominant performance gain (AUC 0.732 to 0.875) and surpassing the single-stage XGBoost baseline reported by Zhou (2025). Version 3 integrates five NOAA meteorological features across 10 major U.S. airports, achieving a final test set AUC of 0.879. FlightSense is deployed as a production AWS MLOps pipeline incorporating live weather ingestion via Lambda, real-time SageMaker inference, an interactive Streamlit dashboard, and an Amazon Bedrock Nova Micro conversational assistant answering natural-language delay queries via a tool-use architecture. --- ### MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference Authors: Ruijie Zhou, Fanxu Meng, Yufei Xu, Tongxuan Liu, Guangming Lu, Muhan Zhang, Wenjie Pei Published: 2026-05-08 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.07363v1 DeepSeek Sparse Attention (DSA) sets the state of the art for fine-grained inference-time sparse attention by introducing a learned token-wise indexer that scores every prefix token and selects the most relevant ones for the main attention. To remain expressive, the indexer uses many query heads (for example, 64 on DeepSeek-V3.2) that share the same selected token set; this multi-head design is precisely what makes the indexer the dominant cost on long contexts. We propose MISA (Mixture of Indexer Sparse Attention), a drop-in replacement for the DSA indexer that treats its indexer heads as a pool of mixture-of-experts. A lightweight router uses cheap block-level statistics to pick a query-dependent subset of only a few active heads, and only those heads run the heavy token-level scoring. This preserves the diversity of the original indexer pool while reducing the per-query cost from scoring every prefix token with every head to scoring it with only a handful of routed heads, plus a negligible router term computed on a small set of pooled keys. We further introduce a hierarchical variant of MISA that uses the routed pass to keep an enlarged candidate set and then re-ranks it with the original DSA indexer to recover the final selected tokens almost exactly. With only eight active heads and no additional training, MISA matches the dense DSA indexer on LongBench across DeepSeek-V3.2 and GLM-5 while running with eight and four times fewer indexer heads respectively, and outperforms HISA on average. It also preserves fully green Needle-in-a-Haystack heatmaps up to a 128K-token context and recovers more than 92% of the tokens selected by the DSA indexer per layer. Our TileLang kernel delivers roughly a 3.82 times speedup over DSA's original indexer kernel on a single NVIDIA H200 GPU. --- ### UniISP: A Unified ISP Framework for Both Human and Machine Vision Authors: Hanxi Li, Yao Cheng, Bo Zhang, Li Zeng Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07359v1 Compared to RGB images, raw sensor data provides a richer representation of information, which is crucial for accurate recognition, particularly under challenging conditions such as low-light environments. The traditional Image Signal Processing (ISP) pipeline generates visually pleasing RGB images for human perception through a series of steps, but some of these operations may adversely impact the information integrity by introducing compression and loss. Furthermore, in computer vision tasks that directly utilize raw camera data, most existing methods integrate minimal ISP processing with downstream networks, yet the resulting images are often difficult to visualize or do not align with human aesthetic preferences. This paper proposes UniISP, a novel ISP framework designed to simultaneously meet the requirements of both human visual perception and computer vision applications. By incorporating a carefully designed Hybrid Attention Module (HAM) and employing supervised learning, the proposed method ensures that the generated images are visually appealing. Additionally, a Feature Adapter module is introduced to effectively propagate informative features from the ISP stage to subsequent downstream networks. Extensive experiments demonstrate that our approach achieves state-of-the-art performance across various scenarios and multiple datasets, proving its generalizability and effectiveness. --- ### GraphReAct: Reasoning and Acting for Multi-step Graph Inference Authors: Xingtong Yu, Zhongwei Kuai, Chang Zhou, Xuanting Xie, Renhe Jiang, Xikun Zhang, Hong Cheng, Xinming Zhang, Yuan Fang Published: 2026-05-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.07357v2 Reasoning-acting frameworks enhance large language models (LLMs) by interleaving reasoning with actions for dynamic information acquisition. However, extending this paradigm to graph learning remains underexplored. Graph data is inherently structured, with information distributed across nodes and edges and encoded through both topology and latent representations. As a result, effective reasoning over graphs requires not only retrieving informative evidence from the graph, but also progressively refining the accumulated context during multi-step inference. In this work, we propose GraphReAct, a graph reasoning-acting framework that enables step-by-step inference over graph-structured data. Specifically, we design a graph-based action space with two complementary retrieval actions: topological retrieval, which captures local structural dependencies, and semantic retrieval, which accesses non-local but relevant evidence in the representation space. These actions dynamically expand the reasoning context. To further support multi-step reasoning, we introduce another type of action, context refinement, which distills and reorganizes accumulated information into a compact representation. By interleaving reasoning with both retrieval and refinement actions, our framework enables a progressive transition from context expansion to compression. Extensive experiments on six benchmark datasets demonstrate that GraphReAct consistently outperforms state-of-the-art methods, validating the effectiveness of reasoning-acting for graph learning. --- ### UniD-Shift: Towards Unified Semantic Segmentation via Interpretable Share-Private Multimodal Decomposition Authors: Shuai Zhang, Zhecheng Shi, Zhuxiao Li, Jing Ou, Tengxi Wang, Yuan Liu, Wufan Zhao Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07356v1 Semantic segmentation of large-scale 3D point clouds is crucial for applications such as autonomous driving and urban digital twins. However, the sparse sampling pattern of LiDAR and the view-dependent geometric distortion in image observations complicate cross-modal alignment and hinder stable fusion. Inspired by the fact that 2D images captured by cameras are representations of the 3D world, we recognize that the features learned from 2D and 3D segmentation share some common semantics, while other aspects remain modality-specific. This insight motivates a unified multimodal framework for joint 2D-3D semantic segmentation. We combine a SAM-based vision encoder with a SPTNet-based geometric encoder to extract complementary semantic and geometric representations. The resulting features from both modalities are explicitly decomposed into shared and private subspaces, where the shared components summarize semantic factors common to both domains, and the private components preserve properties that are unique to each modality. A lightweight attention-based fusion module aggregates the shared features into a consistent cross-modal representation, and a regularized training objective ensures both semantic alignment and subspace independence. Experiments on the SemanticKITTI and nuScenes benchmarks demonstrate consistent improvements in segmentation accuracy over representative multimodal baselines, accompanied by competitive computational efficiency. Cross-domain evaluation on nuScenes USA-Singapore shows stable performance under distribution shifts, demonstrating strong generalization. The implementation code is publicly available at: https://github.com/shuaizhang69/UniD-Shift. --- ### TTF: Temporal Token Fusion for Efficient Video-Language Model Authors: Simin Huo, Ning LI Published: 2026-05-08 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.07355v1 Video-language models (VLMs) face rapid inference costs as visual token counts scale with video length. For example, 32 frames at $448{\times}448$ resolution already yield >8,000 visual tokens in Qwen3-VL, making LLM prefill the dominant throughput bottleneck. Existing methods often rely on global similarity or attention-guided compression, incurring offsets to their gains. We propose \textbf{Temporal Token Fusion (TTF)}, a training-free, plug-and-play pre-LLM token compression framework that exploits structured temporal redundancy in video. TTF automatically selects an anchor frame, then for each subsequent frame, performs a local window similarity search (e.g.,$3\times 3$), fusing tokens that exceed a threshold. The compressed sequence maintains positional consistency across both prefill and decoding through coordinate realignment, enabling seamless integration with existing VLM pipelines. On Qwen3-VL-8B with threshold t=0.70, TTF removes about 67\% of visual tokens while retaining 99.5\% of the baseline accuracy and introducing only ${\approx}0.16$\,GFLOPs of matching overhead. Overall, TTF offers a practical, efficient solution for video understanding. The code is available at \href{https://github.com/Cominder/ttf}{https://github.com/Cominder/ttf} --- ### Task-Oriented Communication for Human Action Understanding via Edge-Cloud Co-Inference Authors: Jingyi Liu, Cheng Yuan, Lijun He, Jun Zhang, Jiawei Shao Published: 2026-05-08 Categories: eess.SP, cs.CV arXiv: https://arxiv.org/abs/2605.07354v1 The expanding application of smart sensing has created a growing demand for the accurate understanding of human action at the network edge. Traditional approaches require massive video data to be transmitted from resource-constrained edge devices to powerful cloud servers, incurring prohibitive uplink bandwidth consumption and unacceptable latency while raising privacy concerns. To overcome these bottlenecks, we propose a task-oriented communication framework for human action understanding (TOAU) through edge-cloud collaboration. Our framework utilizes a monocular pose estimator to extract continuous joint coordinates from raw videos, followed by a vector quantized variational autoencoder (VQ-VAE) to convert these coordinates into discrete motion tokens. Consequently, only a compact sequence of codebook indices is transmitted over the network, consuming as few as 9 bits per frame and avoiding privacy leakages. At the cloud server, a lightweight projector aligns these motion tokens with the embedding space of a large vision-language model (VLM) to facilitate complex action understanding, which is trained with an efficient instruction tuning paradigm. Comprehensive evaluations on three benchmarks demonstrate that our TOAU system reduces the transmission payload to approximately 1\% and the system latency to around 20\% compared to video codec-based solutions, while delivering comparable action understanding accuracy. --- ### Confidence-Aware Alignment Makes Reasoning LLMs More Reliable Authors: Kejia Chen, Jiawen Zhang, Yihong Wu, Kewei Gao, Jian Lou, Zunlei Feng, Mingli Song, Ruoxi Jia Published: 2026-05-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.07353v1 Large reasoning models often reach correct answers through flawed intermediate steps, creating a gap between final accuracy and reasoning reliability. Existing alignment strategies address this with external verifiers or massive sampling, limiting scalability. In this work, we introduce CASPO (Confidence-Aware Step-wise Preference Optimization), a framework that aligns token-level confidence with step-wise logical correctness through iterative Direct Preference Optimization, without training a separate reward model. During inference, we propose Confidence-aware Thought (CaT), which leverages this calibrated confidence to dynamically prune uncertain reasoning branches with negligible O(V) latency. Experiments across ten benchmarks and multiple model families show that CASPO consistently improves reasoning reliability and inference efficiency. CASPO scales to Qwen3-8B-Base and surpasses tree-search baselines on AIME'24 and AIME'25 without using reward-model data. We also release a step-wise dataset with confidence annotations to support fine-grained analysis of reasoning reliability. Code is available at https://github.com/Thecommonirin/CASPO. --- ### Disambiguating 2D-3D Correspondences in Gaussian Splatting-based Feature Fields for Visual Localization Authors: Miso Lee, Sangeek Hyun, Yerim Jeon, Jae-Pil Heo Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07351v1 While Gaussian Splatting-based Feature Fields (GSFFs) have shown promise for visual localization, this paper highlights that photometrically optimized GSFFs are inherently ill-suited for 2D-3D matching. The volumetric extent of each Gaussian induces many-to-one pixel-to-point mappings that destabilize PnP-based pose estimation, while photometric optimization gives rise to superfluous Gaussians devoid of multi-view consistency. To address these issues, we propose SplitGS-Loc, a localization-specialized GSFFs construction framework that disambiguates 2D-3D correspondences by exploiting Gaussian attributes. Our key design, Mixture-of-Gaussians-based splitting, decomposes each Gaussian into smaller Gaussians, replacing ambiguous many-to-one with precise one-to-one correspondences. In parallel, we exploit composition weights from GS rasterization to select Gaussians that significantly and consistently contribute across multiple views and aggregate discriminative features through strong pixel-Gaussian associations, enforcing multi-view consistency. The resulting compact yet discriminative feature fields enable stable PnP convergence, achieving state-of-the-art performance on localization benchmarks. Extensive experiments validate that SplitGS-Loc extends the utility of photometric GSFFs to accurate and efficient localization by exploiting Gaussian attributes, without per-scene training or iterative pose refinement. --- ### SoLAR: Error-Resilient Streamable Long-Horizon Free-Viewpoint Video Reconstruction with Anchor Activation and Latent Recalibration Authors: Haotian Zhang, Xu Mo, Yixin Yu, Guanhua Zhu, Jian Xue, Tongda Xu, Yan Wang, Jiaqi Zhang, Siwei Ma, Wen Gao Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07346v1 Free-Viewpoint Video (FVV) has emerged as a cornerstone of next-generation immersive media systems and attracted widespread attention. Previous methods primarily focus on short video sequences and suffer from significant performance degradation when processing long-horizon free-viewpoint video (LFVV). Motivated by bit allocation theory, we analyze dynamic-anchor-based volumetric video representation within a rate-distortion optimization framework and propose \textbf{SoLAR}, which is the first error-resilient streamable FVV framework that maintains stable reconstruction quality on long sequences without requiring group-of-pictures partitioning. We propose the Anchor Activation Dynamics (AAD), which enables dynamic anchors to model non-rigid transformations by dynamically activating informative anchors and suppressing redundant ones. Furthermore, we introduce Latent Discrepancy Aware Recalibration (LaDAR), which is a mechanism to identify discrepancies between latent representations and recalibrate the correspondences encoded in the network, effectively mitigating error propagation in LFVV without compromising real-time performance or storage compactness. Extensive experiments demonstrate that \textbf{SoLAR} achieves state-of-the-art reconstruction performance while maintaining minimum storage overhead, which provides a new direction for LFVV reconstruction and advances the practical deployment of immersive systems. Demo free-viewpoint videos are provided in the supplementary material. --- ### Mean-Pooled Cosine Similarity is Not Length-Invariant: Theory and Cross-Domain Evidence for a Length-Invariant Alternative Authors: Sibayan Mitra, Dhruv Kumar Published: 2026-05-08 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2605.07345v1 Mean-pooled cosine similarity is the default metric for comparing neural representations across languages, modalities, and tasks. We establish that this metric is not length-invariant: under the anisotropy that characterizes modern transformer representations, mean-pooled cosine grows monotonically in sequence length, independent of representational content. Empirically, on HumanEvalPack across four code LLMs, the length ratio alone explains $R^2 = 0.52$--$0.75$ of cross-language "Python proximity," while AST depth and shared-token fraction add less than 3% of explained variance beyond length. Substituting Centered Kernel Alignment (CKA) reduces explained variance by 83% and reverses the sign of the length coefficient ($β_{\mathrm{len}}: +0.86 \to -0.37$). The same pattern holds in Mistral-7B on parallel WMT pairs ($R^2 = 0.23$ EN-FR, $R^2 = 0.33$ EN-DE for cosine; $R^2 < 0.01$ for CKA). In CLIP ViT-B/32, mean-pooling reduces the length effect relative to EOS-pooling ($R^2: 0.21 \to {<}0.01$), as predicted by the theory's dependence on anisotropy. We argue that length-invariant metrics such as CKA should be the default for cross-representation comparisons, and that recent claims of cross-lingual representational convergence built on mean-pooled cosine warrant re-examination. --- ### Mage: Multi-Axis Evaluation of LLM-Generated Executable Game Scenes Beyond Compile-Pass Rate Authors: Hugh Xuechen Liu, Kıvanç Tatar Published: 2026-05-08 Categories: cs.LG, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2605.07342v1 Compile-pass rate is the dominant evaluation signal for LLM code generation, yet for multi-component domain-specific artifacts it can be actively misleading. We demonstrate this on executable game scene synthesis with a four-axis evaluation protocol (named `Mage') -- compile success, runtime success, structural fidelity, and mechanism adherence -- applied to 858 generation attempts across four open-weight LLMs (7B--30B), 26~hand-crafted Unity goal pattern playable concepts, and two automatically extracted IR granularity levels. Direct NL-to-C\# generation achieves the highest runtime-pass rate (43\% mean) yet produces structurally vacuous scenes (mechanism $F_1 \approx 0.12$). Structural IR conditioning halves the runtime rate but recovers domain-faithful structure ($F_1$ up to 1.00). Within IR conditioning, behavior-only and full-scene granularity are statistically indistinguishable (McNemar $p = 1.0$), indicating input-level granularity saturation. These results show that compile rate is anti-correlated with functional correctness in this domain and that multi-axis evaluation is necessary to detect the divergence. We release the benchmark, replay logs, and per-record metrics for independent verification. --- ### Tools as Continuous Flow for Evolving Agentic Reasoning Authors: Tairan Huang, Siyu Shang, Qiang Chen, Xiu Su, Yi Chen Published: 2026-05-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.07339v2 Large Language Models (LLMs) have demonstrated remarkable capabilities in orchestrating tools for reasoning tasks. However, existing methods rely on a step-wise paradigm that lacks a global perspective, which causes error accumulation over long horizons and restricts generalization to unseen tools. To overcome these limitations, we propose Tools as Continuous Flow for Evolving Agentic Reasoning (FlowAgent), which reconceptualizes tool chaining as continuous trajectory generation within a semantic space. To systematically evaluate this paradigm, we introduce the first plan-level closed-loop benchmark dedicated to plan-level agentic reasoning in dynamic real-world environments. Specifically, the proposed FlowAgent leverages conditional flow matching to generate continuous latent trajectories, providing a global planning perspective to ensure coherent and robust tool execution. Theoretically, we establish formal bounds on utility convergence and prove that our continuous formulation fundamentally guarantees robust generalization and error attenuation. Empirical evaluations show that FlowAgent achieves superior robustness and adaptability in long-horizon reasoning tasks. --- ### Tools as Continuous Flow for Evolving Agentic Reasoning Authors: Tairan Huang, Siyu Shang, Qiang Chen, Xiu Su, Yi Chen Published: 2026-05-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.07339v1 Large Language Models (LLMs) have demonstrated remarkable capabilities in orchestrating tools for reasoning tasks. However, existing methods rely on a step-wise paradigm that lacks a global perspective, which causes error accumulation over long horizons and restricts generalization to unseen tools. To overcome these limitations, we propose Tools as Continuous Flow for Evolving Agentic Reasoning (FlowAgent), which reconceptualizes tool chaining as continuous trajectory generation within a semantic space. To systematically evaluate this paradigm, we introduce the first plan-level closed-loop benchmark dedicated to plan-level agentic reasoning in dynamic real-world environments. Specifically, the proposed FlowAgent leverages conditional flow matching to generate continuous latent trajectories, providing a global planning perspective to ensure coherent and robust tool execution. Theoretically, we establish formal bounds on utility convergence and prove that our continuous formulation fundamentally guarantees robust generalization and error attenuation. Empirical evaluations show that FlowAgent achieves superior robustness and adaptability in long-horizon reasoning tasks. --- ### ShellfishNet: A Domain-Specific Benchmark for Visual Recognition of Marine Molluscs Authors: Ziheng Zhou, Yang Wang, Nan Wang, Chengliang Wu, Jun Yan Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07338v2 Reliable visual recognition of shellfish is important for ecological monitoring and aquaculture, but real-world imagery exhibits substantial variation in acquisition conditions, appearance, and posture. Existing marine vision resources cover diverse taxa and imaging settings, but fine-grained mollusc recognition remains underexplored in a group-disjoint benchmark that jointly evaluates class prediction, representation-space retrieval, and condition sensitivity. To address this gap, we introduce ShellfishNet, a provenance-aware, group-disjoint benchmark of 40,130 images from 100 marine mollusc classes, organized into 30,879 acquisition or observation groups. Rather than treating recognition as a single clean-accuracy leaderboard, ShellfishNet evaluates model configurations across classification, image-to-image retrieval, taxonomic error structure, and controlled condition sensitivity. We benchmark representative convolutional, transformer-based, hybrid and efficient, self-supervised, and vision-language models on classification and retrieval. Separately, we evaluate open image-captioning and vision-language models on the captioning subset and assess selected fixed checkpoints under controlled generic and marine-inspired synthetic transformations. Across the benchmark, image-level and class-aware classification largely agree, whereas retrieval provides complementary model rankings and taxonomy projection reveals that a non-negligible subset of exact-class errors remains correct at higher taxonomic ranks. Selected fixed checkpoints with similar clean accuracy can also exhibit distinct sensitivity profiles under controlled transformations. These results show that model comparison on shellfish imagery should account for both task and image condition, rather than rely on a single clean classification score. The dataset and evaluation protocols will be publicly released. --- ### ShellfishNet: A Domain-Specific Benchmark for Visual Recognition of Marine Molluscs Authors: Ziheng Zhou, Yang Wang, Nan Wang, Chengliang Wu, Jun Yan Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07338v1 The decline of global shellfish biodiversity poses a severe threat to coastal ecosystems. Although artificial intelligence (AI) technologies show potential for automated ecological monitoring, existing marine benthic datasets often lack adaptation to the complexities of real underwater environments (e.g., variable lighting conditions and diverse species postures), posing challenges for the robust generalization of vision models in practical ecological monitoring. To address this problem, we construct ShellfishNet, a comprehensive image benchmark dataset designed specifically for real-world ecological monitoring constraints. Comprising 8,691 images across 32 taxa, this dataset includes a curated subset annotated with descriptive captions. It is constructed through field photography and web scraping, encompassing samples from complex real-world environments. Based on this benchmark, we systematically evaluate 80 representative neural network models, including Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), State Space Models (SSMs), and Self-Supervised Learning (SSL) methods. Furthermore, we evaluate the performance of fine-grained visual categorization (FGVC) models and investigate the image captioning capabilities of several mainstream multimodal large language models (MLLMs). Meanwhile, we introduce image corruption benchmark tests to simulate common underwater degradation scenarios (turbidity, severe weather) and assess the robustness of vision models, enabling trustworthy decisions on ecological protection in the wild. ShellfishNet is dedicated to providing a data foundation and a model-evaluation benchmark for the intelligent monitoring of benthic organisms. --- ### CellScientist: From Execution Feedback to Auditable Model-Revision Trajectories for Cellular Perturbation Prediction Authors: Mengran Li, Bo Li, Jiaying Wang, Wenbin Xing, Chengyang Zhang, Jinlin Wu, Zhen Lei, Jiebo Luo, Stan Z. Li, Zelin Zang Published: 2026-05-08 Categories: cs.LG, cs.SE arXiv: https://arxiv.org/abs/2605.07335v2 Cellular perturbation-response modeling requires coordinated choices of representations, fusion mechanisms, objectives, and training procedures. Large language models (LLMs) can propose executable candidates, but unconstrained revision can produce invalid implementations, change task semantics, or discard useful components. We present CellScientist, a protocol-constrained workflow that converts execution and validation feedback into auditable model-revision trajectories. It records design states and outcomes, routes discrepancies to specific components, and applies local revisions under a fixed task contract. A matched-budget study fixes the candidate language, predictor, fitting, and evaluator: structured revision finds better held-out predictors at small budgets under two LLM backbones. Operational audits link history to fewer repeated proposals, contract checks to contained violations, and discrepancy routing to targeted repairs. Refits of two frozen designs on an independently acquired cohort evaluate external predictive utility. Open-workflow trajectories retain improvements, regressions, and failures, while transcriptomic and single-cell searches extend application to additional response spaces. CellScientist produces both a selected predictor and an inspectable record of its development. Project page: https://limengran98.github.io/CellScientist/. --- ### CellScientist: Dual-Space Hierarchical Orchestration for Closed-Loop Refinement of Virtual Cell Models Authors: Mengran Li, Bo Li, Jiaying Wang, Wenbin Xing, Yixuan Dong, Chengyang Zhang, Hongliang Zhang, Yuzhong Peng, Jinlin Wu, Bob Zhang, Bingo Wing-Kuen Ling, Fuji Yang et al. Published: 2026-05-08 Categories: cs.LG, cs.SE arXiv: https://arxiv.org/abs/2605.07335v1 Virtual Cell Modeling (VCM) requires models that not only predict perturbation responses, but also support targeted revision when predictions fail. Current LLM-assisted modeling workflows face a refinement-routing problem: prediction discrepancies are observed through executable implementations, but the relevant revision may involve the modeling assumption, representation design, implementation, or task constraint. Without structured feedback propagation across these levels, iterative refinement may repair code while failing to revise the assumption responsible for the discrepancy. We propose CellScientist, a dual-space hierarchical framework that couples a high-level hypothesis space with a low-level executable implementation space. CellScientist represents modeling decisions as structured states, realizes them as admissible programs under task and interface constraints, and routes execution discrepancies back to targeted hypothesis or implementation updates. This enables a closed Hypothesis -> Implementation -> Hypothesis loop where failures become structured signals for model refinement rather than debugging events. Across morphology and transcriptomic benchmarks, with additional single-cell perturbation evaluations, the final executable models selected by CellScientist improve over reference baselines under fixed split and evaluation protocols, while the workflow produces auditable refinement traces. --- ### RCoT-Seg: Reinforced Chain-of-Thought for Video Reasoning and Segmentation Authors: Junwei Wen, Deshui Miao, Guangming Lu, Xin Li, Wenjie Pei Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07334v1 Video Reasoning Segmentation (VRS) aims to segment target objects in videos based on implicit instructions that convey human intent and temporal logic. Existing MLLM-based methods predict masks with a (SEG) token after selecting frames via simple sampling or an auxiliary MLLM, where limited supervision and frame-language similarity rules often yield narrow-scope keyframe choices that weaken holistic temporal understanding and lead to brittle localization in complex multi-object scenes. To address these issues, we introduce RCoT-Seg, a video-of-thought framework that factorizes VRS into temporal video reasoning (TVR) and keyframe target perception (KTP), explicitly separating temporal reasoning from spatial perception. Specifically, in the TVR stage, an agentic keyframe selection module, initialized with a curated CoT-start corpus and refined by GRPO under task-aligned rewards, is proposed to generate and reselect the keyframe through self-evaluation, strengthening moment localization and temporal reasoning. In the KTP stage, RCoT-Seg performs high-resolution segmentation on the selected frame and propagates masks with SAM2-based methods across the sequence, replacing heuristic sampling and external selectors while improving spatial precision and inter-frame consistency. Extensive experimental results demonstrate that the proposed RCoT-Seg achieves favorable performance against the state-of-the-art methods. The code and models will be publicly released at https://github.com/Victor-wjw/RCoT-Seg. --- ### Beyond Linear Attention: Softmax Transformers Implement In-Context Reinforcement Learning Authors: Zixuan Xie, Xinyu Liu, Claire Chen, Shuze Daniel Liu, Rohan Chandra, Shangtong Zhang Published: 2026-05-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.07333v2 In-context reinforcement learning (ICRL) studies agents that, after pretraining, adapt to new tasks by conditioning on additional context without parameter updates. Existing theoretical analyses of ICRL largely rely on linear attention, which replaces the softmax function in the standard attention with an identity mapping. This paper provides the first theoretical understanding of ICRL without making the unrealistic linear attention simplification. In particular, we consider the standard softmax attention used in practice. We show that, with certain parameters, the layerwise forward pass of a Transformer with such softmax attention is equivalent to iterative updates of a weighted softmax temporal difference (TD) learning algorithm. Here, weighted softmax TD is a new RL algorithm that performs policy evaluation in kernel space and adopts both linear TD and tabular TD as special cases. We also prove that under a certain contraction condition, the policy evaluation error decays as the number of layers grows, with the identified parameters above. Finally, we prove that those parameters are a global minimizer of a pretraining loss, explaining their emergence in our numerical experiments. --- ### Rethinking Importance Sampling in LLM Policy Optimization: A Cumulative Token Perspective Authors: Yuheng Zhang, Chenlu Ye, Shuowei Jin, Changlong Yu, Wei Xiong, Saurabh Sahu, Nan Jiang Published: 2026-05-08 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.07331v1 Reinforcement learning, including reinforcement learning with verifiable rewards (RLVR), has emerged as a powerful approach for LLM post-training. Central to these approaches is the design of the importance sampling (IS) ratio used in off-policy policy-gradient estimation. Existing methods face a fundamental bias-variance dilemma: token-level IS ratios, as adopted by PPO (Schulman et al., 2017) and GRPO (Shao et al., 2024), introduce bias by ignoring prefix state distribution mismatch; full sequence ratios provide exact trajectory-level correction but suffer from high variance due to the multiplicative accumulation of per-token ratios, while GSPO (Zheng et al., 2025) improves numerical stability via length normalization at the cost of deviating from the exact full-sequence IS correction. In this work, we identify the cumulative token IS ratio, the product of per-token ratios up to position $t$, as a theoretically principled solution to this dilemma. We prove that, under the token-level policy-gradient formulation, this ratio provides an unbiased prefix correction for each token-level gradient term and has strictly lower variance than the full sequence ratio. Building on this insight, we propose CTPO (Cumulative Token Policy Optimization), which combines the cumulative token IS ratio with position-adaptive clipping that scales log-space clip bounds according to the natural $\sqrt{t}$ growth of the cumulative log-ratio. This yields more consistent regularization across token positions. We implement and evaluate CTPO in the tool-integrated reasoning setting on several challenging mathematical reasoning benchmarks, achieving the best average performance across both model scales compared with strong GRPO and GSPO baselines. Code will be available at https://github.com/horizon-llm/CTPO. --- ### SparseRL-Sync: Lossless Weight Synchronization with ~100x Less Communication Authors: Lucas Hu, Ranchi Zhao, Isaac Zhu, Zach Zhang, Hscos Zhang, Hugh Yin, Jason Zhao Published: 2026-05-08 Categories: cs.LG, cs.AI, cs.DC arXiv: https://arxiv.org/abs/2605.07330v1 In large-scale reinforcement learning (RL) systems with decoupled Trainer-Rollout execution, the Trainer must regularly synchronize policy weights to the Rollout side to limit policy staleness. When inter-node bandwidth is abundant, such synchronization is usually only a small fraction of end-to-end cost. As model size grows, however, the communication demand rises rapidly. In bandwidth-constrained or network-variable deployments -- for example, cross-datacenter or cross-cluster settings, heterogeneous resource pools, and online RL -- weight synchronization can become a dominant bottleneck for throughput and tail latency. We observe that, in mainstream large-model RL training, the locations where parameters actually change are highly sparse at the element level (often 99%+ sparsity). Building on this observation, we propose and implement SparseRL-Sync, which replaces full-weight transfers with a lossless sparse update payload (indices and values) that can be exactly reconstructed on the inference side, thereby preserving 100% fidelity. Under a simplified cost model, sparse synchronization reduces the per-update communication volume from S to approximately S/X; with 99% sparsity (X ~ 100), this yields about a 100x reduction in transmitted data. Combined with appropriate bucketing, SparseRL-Sync also reduces launch and control-plane overhead, significantly improving scalability and end-to-end efficiency in bandwidth-limited and highly asynchronous RL settings. --- ### GC-ART: Global Learnable Second-Order Rational Tone Curves for Illumination Robustness Authors: Wei Huang, Joyce Huang Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07329v1 We introduce GC-ART (Global Curve Adaptive Rational Tone-mapping), a lightweight differentiable pre-processing module for robust image classification. GC-ART predicts an endpoint-pinned rational tone curve from per-channel soft histograms using a 643-parameter MLP, then applies the curve pointwise before the classifier. The module is trained end-to-end with cross-entropy and a soft monotonicity penalty. On CIFAR-10 with a CIFAR-style ResNet-18, GC-ART matches clean accuracy with the unenhanced baseline and other learned enhancers, improves over the baseline on multiplicative darkening, and achieves the best learned-method result on contrast corruption (48.45% vs. 46.27% for the baseline and 47.13% for Zero-DCE++). These results suggest that histogram-conditioned rational curves can learn useful global tone corrections, including contrast-expanding behavior, while preserving edge locations by construction through pointwise mapping. GC-ART also uses substantially fewer FLOPs than convolutional learned enhancers at 32 x 32. The current hyperparameters are untuned, leaving room for systematic improvement. --- ### Teacher-Feature Drifting: One-Step Diffusion Distillation with Pretrained Diffusion Representations Authors: Yuan Zhang, Chenyi Li, Haodong Yu, Guoqing Ma, Jiajun Zha, Yuanming Yang, Bo Wang, Wei Tang, Wenbo Li, Haoyang Huang, Nan Duan Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07327v2 Sampling from pretrained diffusion and flow-matching models typically requires many forward passes to generate diverse and high-fidelity images. Existing distillation methods often rely on multiple auxiliary networks, carefully designed training stages, or complex optimization pipelines. In this work, we revisit the recently proposed Drifting Model objective and show that a single drifting loss can be directly used to simplify one step distillation. A key observation is that the pretrained diffusion teacher itself already provides a strong representation space. Unlike the original Drifting Model, which relies on an additional pretrained feature extractor, we use intermediate hidden states of the pretrained teacher model as the feature representation. This removes the need for training or introducing an extra representation network while preserving a semantically meaningful feature geometry for drifting. Furthermore, we introduce a lightweight mode coverage loss to mitigate mode collapse during distillation and encourage the student generator to cover diverse teacher-supported regions. Extensive experiments on ImageNet and SDXL demonstrate that our method achieves efficient one step generation with competitive image quality and diversity, achieving FID scores of 1.58 on ImageNet-64$\times$64 and 18.4 on SDXL, while substantially simplifying the overall distillation framework. --- ### Teacher-Feature Drifting: One-Step Diffusion Distillation with Pretrained Diffusion Representations Authors: Yuan Zhang, Chenyi Li, Guoqing Ma, Jiajun Zha, Yuanming Yang, Bo Wang, Wei Tang, Wenbo Li, Haoyang Huang, Nan Duan Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07327v1 Sampling from pretrained diffusion and flow-matching models typically requires many forward passes to generate diverse and high-fidelity images. Existing distillation methods often rely on multiple auxiliary networks, carefully designed training stages, or complex optimization pipelines. In this work, we revisit the recently proposed Drifting Model objective and show that a single drifting loss can be directly used to simplify one step distillation. A key observation is that the pretrained diffusion teacher itself already provides a strong representation space. Unlike the original Drifting Model, which relies on an additional pretrained feature extractor, we use intermediate hidden states of the pretrained teacher model as the feature representation. This removes the need for training or introducing an extra representation network while preserving a semantically meaningful feature geometry for drifting. Furthermore, we introduce a lightweight mode coverage loss to mitigate mode collapse during distillation and encourage the student generator to cover diverse teacher-supported regions. Extensive experiments on ImageNet and SDXL demonstrate that our method achieves efficient one step generation with competitive image quality and diversity, achieving FID scores of 1.58 on ImageNet-64$\times$64 and 18.4 on SDXL, while substantially simplifying the overall distillation framework. --- ### GEM: Generating LiDAR World Model via Deformable Mamba Authors: Yang Wu, Zhaojiang Liu, Qiang Meng, Youquan Liu, Renliang Weng, Jianjun Qian, Jian Yang, Jin Xie Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07326v2 World models, which simulate environmental dynamics and generate sensor observations, are gaining increasing attention in autonomous driving. However, progress in LiDAR-based world models has lagged behind those built on camera videos or occupancy data, primarily due to two core challenges: the inherent disorder of LiDAR point clouds and the difficulty of distinguishing dynamic objects from static structures. To address these issues, we propose GEM: a Generative LiDAR world model that leverages deformable mamba architecture, significantly improving fidelity and imaginative capability. Specifically, leveraging the structural similarity between sequential laser scanning and Mamba's processing mechanism, we first tokenize LiDAR sweeps into compact representations via a custom LiDAR scene tokenizer. After unsupervised disentanglement of tokenized features via a dynamic-static separator, a tri-path deformable Mamba is introduced to perform selective scanning and adaptive gating fusion over the disentangled features, leading to enhanced spatial-temporal understanding of the world evolution. Optionally, a planner and a BEV layout controller can be integrated to explore the model's capability for autonomous rollout and its potential to generate ``what-if" scenarios. Extensive experiments show that GEM achieves state-of-the-art performances across diverse benchmarks and evaluation settings, demonstrating its superiority and effectiveness. Project page: https://github.com/wuyang98/GEM. --- ### GEM: Generating LiDAR World Model via Deformable Mamba Authors: Yang Wu, Zhaojiang Liu, Qiang Meng, Youquan Liu, Renliang Weng, Jianjun Qian, Jian Yang, Jin Xie Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07326v1 World models, which simulate environmental dynamics and generate sensor observations, are gaining increasing attention in autonomous driving. However, progress in LiDAR-based world models has lagged behind those built on camera videos or occupancy data, primarily due to two core challenges: the inherent disorder of LiDAR point clouds and the difficulty of distinguishing dynamic objects from static structures. To address these issues, we propose GEM: a Generative LiDAR world model that leverages deformable mamba architecture, significantly improving fidelity and imaginative capability. Specifically, leveraging the structural similarity between sequential laser scanning and Mamba's processing mechanism, we first tokenize LiDAR sweeps into compact representations via a custom LiDAR scene tokenizer. After unsupervised disentanglement of tokenized features via a dynamic-static separator, a tri-path deformable Mamba is introduced to perform selective scanning and adaptive gating fusion over the disentangled features, leading to enhanced spatial-temporal understanding of the world evolution. Optionally, a planner and a BEV layout controller can be integrated to explore the model's capability for autonomous rollout and its potential to generate ``what-if" scenarios. Extensive experiments show that GEM achieves state-of-the-art performances across diverse benchmarks and evaluation settings, demonstrating its superiority and effectiveness. Project page: https://github.com/wuyang98/GEM. --- ### CSR: Infinite-Horizon Real-Time Policies with Massive Cached State Representations Authors: Robin Karlsson, Go Suzui Published: 2026-05-08 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2605.07325v1 Deploying massive large language models (LLMs) as continuous cognitive engines for robotics is bottlenecked by the time-to-first-token (TTFT) latency required to process extensive state histories. Existing solutions like RAG or sliding windows compromise global context or incur prohibitive re-computation costs. We formalize the optimal task structure for minimizing latency and theoretically prove that prefix stability, incremental extensibility, and asynchronous state reconciliation are necessary conditions for real-time performance. Building on these proofs, we introduce the Cached State Representation (CSR) framework as the practical instantiation of these properties, ensuring optimal KV-cache reuse. To sustain these properties over infinite horizons, we further propose an Asynchronous State Reconciliation (ASR) algorithm that offloads state memory eviction to a parallel computational resource to eliminate latency spikes. On a physical robot wirelessly connected to an on-premise GPU server, CSR achieves a 26-fold latency reduction (14.67s to 0.56s) for 120K token contexts with a 235B parameter model compared to a standard baseline. On an embodied AI benchmark, we achieve SOTA recall (0.836 vs. 0.459) while maintaining RAG-level latency. ASR is validated to sustain bounded, spike-free TTFT over 10 eviction cycles in continuous real-world operation. Together, CSR and ASR enable massive LLMs to function as continuously operating, high-frequency (> 2 Hz) embodied policies. --- ### Activation Differences Reveal Backdoors: A Comparison of SAE Architectures Authors: Sachin Kumar Published: 2026-05-08 Categories: cs.CL, cs.AI, cs.CR, cs.LG arXiv: https://arxiv.org/abs/2605.07324v1 Backdoor attacks on language models pose a significant threat to AI safety, where models behave normally on most inputs but exhibit harmful behavior when triggered by specific patterns. Detecting such backdoors through mechanistic interpretability remains an open challenge. We investigate two sparse autoencoder architectures -- Crosscoders and Differential SAEs (Diff-SAE) -- for isolating backdoor-related features in fine-tuned models. Using a controlled SQL injection backdoor triggered by year-based context ("2024" triggers vulnerable code, "2023" triggers safe code), we evaluate both approaches across LoRA and full-rank fine-tuning regimes on SmolLM2-360M. We find that Diff-SAE consistently and substantially outperforms Crosscoders for backdoor isolation. Diff-SAE achieves a Backdoor Isolation Score (BIS) of 0.40 with perfect precision (1.0) and zero false positive rate across most experimental conditions, while Crosscoders fail almost entirely with BIS below 0.02 in most cases. This performance gap holds across multiple transformer layers (14, 18, 22, 26) and both fine-tuning regimes, with full-rank fine-tuning producing particularly clean backdoor signals. Our results suggest that backdoors manifest as directional activation shifts rather than sparse feature activations, making difference-based representations fundamentally more effective for detection. These findings have important implications for AI safety monitoring and the development of interpretability tools for detecting model manipulation. --- ### Discovering Ordinary Differential Equations with LLM-Based Qualitative and Quantitative Evaluation Authors: Sum Kyun Song, Bong Gyun Shin, Jae Yong Lee Published: 2026-05-08 Categories: cs.AI, cs.LG, cs.NE, cs.SC arXiv: https://arxiv.org/abs/2605.07323v2 Discovering governing differential equations from observational data is a fundamental challenge in scientific machine learning. Existing symbolic regression approaches rely primarily on quantitative metrics; however, real-world differential equation modeling also requires incorporating domain knowledge to ensure physical plausibility. To address this gap, we propose DoLQ, a method for discovering ordinary differential equations with LLM-based qualitative and quantitative evaluation. DoLQ employs a multi-agent architecture: a Sampler Agent proposes dynamic system candidates, a Parameter Optimizer refines equations for accuracy, and a Scientist Agent leverages an LLM to conduct both qualitative and quantitative evaluations and synthesize their results to iteratively guide the search. Experiments on multi-dimensional ordinary differential equation benchmarks demonstrate that DoLQ achieves superior performance compared to existing methods, not only attaining higher success rates but also more accurately recovering the correct symbolic terms of ground truth equations. Our code is available at https://github.com/Bon99yun/DoLQ. --- ### Generative Modeling with Flux Matching Authors: Peter Pao-Huang, Xiaojie Qiu, Stefano Ermon Published: 2026-05-08 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.07319v1 We introduce Flux Matching, a new paradigm for generative modeling that generalizes existing score-based models to a broader family of vector fields that need not be conservative. Rather than requiring the model to equal the data score, the Flux Matching objective imposes a weaker condition that admits infinitely many vector fields whose stationary distribution is the data. This flexibility enables a class of generative models that cannot be learned under score matching, in which inductive biases, structural priors, and properties of the dynamics can be directly imposed or optimized. We show that Flux Matching performs strongly on high-dimensional image datasets and, more importantly, that our added freedom unlocks a range of applications including faster sampling, interpretable and mechanistic models, and dynamics that encode directed dependencies between variables. More broadly, Flux Matching opens a new dimension in generative modeling by turning the vector field itself into a design choice rather than a fixed target. Code is available at https://github.com/peterpaohuang/flux_matching. --- ### Amortized-Precision Quantization for Early-Exit Vision Transformers Authors: Rui Fang, Hsi-Wen Chen, Ming-Syan Chen Published: 2026-05-08 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.07317v1 Vision Transformers (ViTs) achieve strong performance across vision tasks, yet their deployment with low-precision early exiting remains fragile. Existing quantization methods assume static full-depth execution, making them unstable when exit decisions are perturbed by quantization noise, which can amplify errors along dynamic inference paths. In this paper, we introduce Amortized-Precision Quantization (APQ), a utilization-aware formulation that accounts for layer-wise stochastic exposure to quantization noise and reveals depth-precision trade-offs. Building on APQ, we propose Mutual Adaptive Quantization with Early Exiting (MAQEE), a bi-level framework that jointly optimizes exit thresholds and bit-widths under explicit risk control to improve inference stability. MAQEE establishes a superior Pareto frontier in the accuracy-efficiency trade-off, reducing BOPs by up to 95% while maintaining accuracy and outperforming strong baselines by up to 20\% across classification, detection, and segmentation tasks. --- ### Implicit Compression Regularization: Concise Reasoning via Internal Shorter Distributions in RL Post-Training Authors: Chen Wang, Hexuan Deng, Yining Zhang, Yuchen Zhang, Jionghao Bai, Zhaochun Li, Ge Lan, Yue Wang Published: 2026-05-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.07316v1 Reinforcement learning with verifiable rewards improves LLM reasoning but often induces overthinking, where models generate unnecessarily long reasoning traces. Existing methods mainly rely on length penalties or early-exit strategies; however, the former may degrade accuracy and induce underthinking, whereas the latter assumes that substantial portions of reasoning traces can be safely truncated. To obtain a compression signal without these limitations, we revisit the training dynamics of existing compression methods. We observe that the length--accuracy correlation is initially negative but continually increases during compression, indicating that shorter responses are initially more likely to be correct but gradually lose this property as the policy moves toward underthinking. Based on this observation, we formalize overthinking: a negative correlation indicates an overthinking regime, while a positive one indicates underthinking. When overthinking, the shortest correct responses are shorter than the group-average response length in expectation, making them natural compression targets already present in on-policy rollouts. We therefore propose \emph{Implicit Compression Regularization} (ICR), an on-policy regularization method whose compression signal comes from a virtual shorter distribution induced by the shortest correct responses in rollout groups, guiding the policy toward concise yet correct trajectories. Training dynamics show that ICR maintains a better length--accuracy correlation during compression, indicating that short responses remain better aligned with correctness instead of drifting toward underthinking. Experiments on three reasoning backbones and multiple mathematical and knowledge-intensive benchmarks show that ICR consistently shortens responses while preserving or improving accuracy, achieving a stronger accuracy--length Pareto frontier. --- ### LaTER: Efficient Test-Time Reasoning via Latent Exploration and Explicit Verification Authors: Xuan Li, Yining Wang, Yuchen Liu, Guanjun Liu, Delai Qiu, Shengping Liu, Jiaen Liang, Wei Huang, Jun Yu, Junnan Zhu Published: 2026-05-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.07315v1 Chain-of-thought (CoT) reasoning improves large language models (LLMs) on difficult tasks, but it also makes inference expensive because every intermediate step must be generated as a discrete token. Latent reasoning reduces visible token generation by propagating continuous states, yet replacing explicit derivations with latent computation can hurt tasks that require symbolic checking. We propose Latent-Then-Explicit Reasoning (LaTER), a two-stage paradigm that first performs bounded exploration in a continuous latent space and then switches to explicit CoT for verification and answer generation. In a training-free instantiation, LaTER projects final-layer hidden states back to the input embedding space, preserves the latent KV cache, and uses entropy and model-native stop-token probes to decide when to switch. We find that strong reasoning models already exhibit structured latent trajectories under this interface. On Qwen3-14B, training-free LaTER reduces total token usage by 16%-32% on several benchmarks while matching or improving accuracy on most of them; for example, it improves AIME 2025 from 70.0% to 73.3% while reducing tokens from 15,730 to 10,661. We further construct Latent-Switch-69K, a supervised corpus that pairs condensed solution intuitions with shortened explicit derivations. Fine-tuning with latent rollout and halting supervision yields additional gains: trained LaTER reaches 80.0% accuracy on AIME 2025, 10.0 points above the standard CoT baseline, while using 33% fewer tokens. Our code, data, and model are available at https://github.com/TioeAre/LaTER. --- ### DCGL: Dual-Channel Graph Learning with Large Language Models for Knowledge-Aware Recommendation Authors: Xinchi Zou, Tongzhenzhi Su, Jianjun Li, Yuan Fu, Chang Liu, Zhiying Deng, Zhiwei Shen Published: 2026-05-08 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.07314v1 Knowledge Graphs (KGs) have proven highly effective for recommendation systems by capturing latent item relationships, while recent integration of Large Language Models (LLMs) has further enhanced semantic understanding and addressed knowledge sparsity issues. Nevertheless, current KG-and-LLM-based methods still face three main limitations: 1) inadequate modeling of implicit semantic relationships beyond explicit KG links; 2) suboptimal single-channel fusion of ID and LLM embeddings, which often leads to signal interference and blurred representations; and 3) insufficient consideration of user-item interaction frequency variations in recommendation strategies. To address these challenges, we propose the Dual-Channel Graph Learning (DCGL) framework, featuring three key innovations: 1) a dual-channel architecture that structurally decouples rich semantic information from user behavioral patterns, preventing early interference; 2) a multi-level contrastive learning mechanism that enhances robustness against KG noise through intra-view contrasts and bridges semantic gaps between channels via inter-view alignment; and 3) a dynamic fusion mechanism that adaptively balances semantic generalization and behavioral specificity based on interaction frequency, resolving the cascading limitation. Extensive experiments on four real-world datasets show that DCGL consistently outperforms state-of-the-art methods, yielding substantial improvements in sparse scenarios while maintaining precision for active users. Our code is available at https://github.com/XinchiZou/DCGL. --- ### When Stored Evidence Stops Being Usable: Scale-Conditioned Evaluation of Agent Memory Authors: Jiaqi Shao, Yiyi Lu, Yunzhen Zhang, Bing Luo Published: 2026-05-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.07313v1 Memory-agent evaluations report fixed-snapshot accuracy or retrieval quality, but these scores do not show whether evidence remains usable as irrelevant sessions (sessions not annotated as task-relevant evidence for the query) accumulate. We present a scale-conditioned evaluation protocol for agent memory under evidence-preserving growth: for each query, task evidence is held fixed while irrelevant sessions are added. The protocol logs agent--memory trajectories and reports four diagnostics: budget-compliant reliability, tail memory-call burden, failure-regime decomposition, and the usable-scale boundary where reliability falls below the target. Applied to LongMemEval and LoCoMo across flat, planar, and hierarchical memory interfaces, the protocol shows reliability loss is not a single phenomenon. On LongMemEval, HippoRAG stays within the two-call budget but loses 16--20 percentage points in budget-compliant reliability as irrelevant sessions are added; LiCoMemory's observed failures depend strongly on the agent, with Qwen3-8B exceeding the budget while Qwen3-32B and Qwen3-235B remain reliable in the tested range. The result supports a framework for making scalable-memory claims conditional on agent, interface, scale range, and interaction budget. --- ### Rethinking Dense Sequential Chains: Reasoning Language Models Can Extract Answers from Sparse, Order-Shuffling Chain-of-Thoughts Authors: Yi-Chang Chen, Feng-Ting Liao, Da-shan Shiu, Hung-yi Lee Published: 2026-05-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.07307v1 Modern reasoning language models generate dense, sequential chain-of-thought traces implicitly assuming that every token contributes and that steps must be consumed in order. We challenge both assumptions through a systematic intervention pipeline--removal, masking, shuffling, and noise injection--applied to model-generated reasoning chains across three models and three benchmarks. Our findings are counterintuitive on three dimensions. Order: Does the sequential order of a reasoning chain matter for answer extraction? No--line-level shuffling reduces accuracy by less than 0.5 pp; word-level shuffling retains 62%-89% accuracy; only token-level shuffling collapses to near zero. Pretrained-only and instruction-tuned variants exhibit near-identical tolerance (78.67% vs. 78.00% under line shuffling), indicating order-independence originates from pretraining rather than reasoning-specific fine-tuning. Dense: Is all the information in a reasoning chain important for answer extraction? No--masking numeric digits collapses accuracy to exactly 0%, while masking alphabetic prose improves accuracy by 4.7 pp. Robustness: Is a reasoning chain that is both order-shuffling and non-dense still robust? Yes--the most aggressively reduced representation (all natural language removed, lines arbitrarily shuffled) still achieves 83% accuracy, and injecting false answers at 3x true-answer frequency leaves accuracy unchanged (83.3%->83.3%), falsifying a frequency-based extraction account. These results establish that answer extraction operates on a sparse, order-insensitive, and structurally robust informational substrate, opening paths toward parallelized and token-efficient reasoning generation. --- ### BioProVLA-Agent: An Affordable, Protocol-Driven, Vision-Enhanced VLA-Enabled Embodied Multi-Agent System with Closed-Loop-Capable Reasoning for Biological Laboratory Manipulation Authors: Zhaohui Du, Zhe Wang, Dongzhan Zhou, Minting Pan, Hongmei Fei, Xiwen Cao, Ting Xiao, Qi Wang, Huanbo Jin, Jiaming Gu, Quan Lu, Zhe Liu Published: 2026-05-08 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2605.07306v3 Biological laboratory automation can reduce repetitive manual work and improve reproducibility, but reliable embodied execution in wet-lab environments remains challenging. Protocols are often unstructured, labware is frequently transparent or reflective, and multi-step procedures require state-aware execution beyond one-shot instruction following. Existing robotic systems often rely on costly hardware, fixed workflows, dedicated instruments, or robotics-oriented interfaces. Here, we introduce BioProVLA-Agent, an affordable, protocol-driven, vision-enhanced embodied multi-agent system enabled by Vision-Language-Action (VLA) models for biological manipulation. The system uses protocols as the task interface and integrates protocol parsing, visual state verification, and embodied execution in a closed-loop workflow. A Tailored LLM Protocol Agent converts protocols into verifiable subtasks; a VLM-RAG Verification Agent assesses readiness and completion using observations, robot states, retrieved knowledge, and success/failure examples; and a VLA Embodied Agent executes verified subtasks through a lightweight policy. To improve robustness under wet-lab visual perturbations, we develop AugSmolVLA, an online augmentation strategy targeting transparent labware, reflections, illumination shifts, and overexposure. We evaluate the system on a hierarchical benchmark covering 15 atomic tasks, 6 composite workflows, and 3 bimanual tasks, including tube loading, sorting, waste disposal, cap twisting, and liquid pouring. Across normal and high-exposure settings, AugSmolVLA improves execution stability over ACT, X-VLA, and the original SmolVLA, especially for precise placement, transparent-object manipulation, composite workflows, and visually degraded scenes. These results suggest a practical route toward accessible, protocol-centered, and verification-capable embodied AI for biological manipulation. --- ### MedAction: Towards Active Multi-turn Clinical Diagnostic LLMs Authors: Hsin-Ling Hsu, Zizheng Wang, Donghua Zhang, Nai-Chia Chen, Jerry Wang, Jun-En Ding, Chia-Hsuan Hsu, Guoan Wang, Feng Liu, Fang-Ming Hung, Chenwei Wu, Liyue Shen Published: 2026-05-08 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.07305v1 Most existing LLM diagnoses are evaluated on static, single-turn settings where complete patient information is provided upfront, an oversimplification of real clinical practice. We study active diagnosis: the real-life clinical process of starting from initial observation, ordering tests, interpreting results, and updating a differential diagnosis across multiple turns. Through systematic analysis, we identify three recurring failure modes in current LLMs: ungrounded test ordering, unreliable diagnostic update, and degraded multi-turn coherence. Together, these failures reveal a core deficit: existing medical training data teaches models to reason from complete information but not to act under evolving, partial evidence. To address this gap, we introduce MedAction, a tree-structured distillation pipeline that synthesizes diverse and high-quality multi-turn diagnostic trajectories via LLM-environment interaction. We propose two knowledge-graph-grounded metrics to filter trajectory quality: Disease Trajectory Consistency (DTC), which tracks whether the model's hypothesis converges toward the correct diagnosis, and Reasoning-Action Consistency (RAC), which verifies that belief updates are driven by gathered evidence. Using this pipeline, we construct MedAction-32K, a dataset of 32,681 trajectories from 2,896 PMC cases. Fine-tuning an 8B model on MedAction-32K achieves state-of-the-art performance among open-source models on both MedR-Bench and our curated MedAction-300-Hard benchmark, pushing the edge for open-source medical LLMs. --- ### Latent Order Bandits Authors: Emil Carlsson, Newton Mwai, Fredrik D. Johansson Published: 2026-05-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.07304v2 Bandit algorithms solve diverse sequential decision-making problems, but are often too sample-inefficient for from-scratch personalization. To substantially reduce exploration times, latent bandit algorithms exploit cross-instance structure implied by discrete latent states, provided that the posterior distribution of rewards and latent states is known and accurate. However, obtaining an accurate model of this structure is difficult, and a small number of latent states may be insufficient to characterize the reward distributions in all problem instances. We propose latent order bandits (LOB), relaxing the assumptions of latent bandits to require only prior knowledge of a \emph{partial order} of action preferences in each state. This allows instances of the same state to vary in reward distributions, as long as the partial order of actions is shared. For example, groups of users on a streaming service may agree on which movie genres are the best but rate experiences on different scales. We give an upper-confidence bound procedure for the LOB problem, applicable to both total and partial latent orders, and give an upper bound on its regret. To improve empirical performance, we propose a posterior-sampling algorithm and show, in a suite of experiments, that both are competitive with full-prior latent bandits when same-state instances share reward parameters, and preferable to them when reward scales differ between instances with the same latent state. --- ### Latent Order Bandits Authors: Emil Carlsson, Newton Mwai, Fredrik D. Johansson Published: 2026-05-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.07304v1 Bandit algorithms solve diverse sequential decision-making problems, but are often too sample-inefficient for from-scratch personalization. To substantially reduce exploration times, latent bandit algorithms exploit cross-instance structure implied by discrete latent states, provided that the posterior distribution of rewards and latent states is known and accurate. However, obtaining an accurate model of this structure is difficult, and a small number of latent states may be insufficient to characterize the reward distributions in all problem instances. We propose latent order bandits (LOB), relaxing the assumptions of latent bandits to require only prior knowledge of a partial order of action preferences in each state. This allows instances of the same state to vary in reward distributions, as long as the partial order of actions is shared. For example, groups of users on a streaming service may agree on which movie genres are the best but rate experiences on different scales. We give an upper-confidence bound procedure for the LOB problem, applicable to both total and partial latent orders, and give an upper bound on its regret. To improve empirical performance, we propose a posterior-sampling algorithm and show, in a suite of experiments, that both are competitive with full-prior latent bandits when same-state instances share reward parameters, and preferable to them when reward scales differ between instances with the same latent state. --- ### Pretraining Induces a Reusable Spectral Basis for Downstream Task Adaptation Authors: Junjie Yu, Yue Wang, Zihan Deng, Yan Zhu, Wenxiao Ma, Quanying Liu Published: 2026-05-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.07302v1 Finetuning pretrained models occurs in a low-dimensional subspace of the full parameter space. Prior work has focused on characterizing this optimization subspace, but largely ignored the complementary question: why do certain directions remain unexplored during finetuning? Are these stable directions irrelevant to downstream tasks, or do they already encode task-relevant structure that requires no further adjustment? Answering this question is central to understanding how pretrained knowledge transfers. Through systematic spectral analysis across vision and language models, we show that the leading singular vectors of pretrained weight matrices remain highly stable under finetuning and are shared across unrelated downstream tasks, revealing that pretraining establishes a reusable spectral coordinate system. Models pretrained on larger datasets exhibit greater spectral stability under distribution shift or task change, directly linking pretraining scale to geometric transferability. Motivated by these findings, we propose a parameter-efficient method that freezes pretrained singular vectors and optimizes only leading spectral coefficients, achieving competitive performance on GLUE with 0.2% trainable parameters. Our results reveal that the stable directions encode transferable structure rather than irrelevant noise: successful pretraining discovers spectral bases that downstream tasks inherit and operate within. --- ### SOM: Structured Opponent Modeling for LLM-based Agents via Structural Causal Model Authors: Shiyue Cao, Pei Xu, Likun Yang, Lei Cui, Xiaotang Chen, Kaiqi Huang Published: 2026-05-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.07301v1 Accurately predicting opponents' behavior from interactions is a fundamental capability for large language model (LLM)-based agents in multi-agent and game-theoretic environments. Existing approaches often entangle opponent modeling with prediction, relying on implicit contextual reasoning and limiting adaptability in dynamic interactions. To this end, we propose Structured Opponent Modeling (SOM), a two-stage opponent modeling framework that distinctly separates opponent model construction and opponent prediction. At the construction stage, SOM employs a Structural Causal Model (SCM), a graph-based formalism for representing dependencies among variables, to capture directed links between opponents' observations and actions, yielding an explicit and structured opponent representation. At the prediction stage, the LLM performs structured reasoning along clear pathways derived from the SCM, improving both prediction accuracy and stability. Extensive experiments on diverse multi-agent benchmarks demonstrate that SOM consistently outperforms state-of-the-art LLM-based reasoning baselines, enabling more accurate and adaptable strategic decision-making in complex and dynamic multi-agent interactions. --- ### EgoPro-Bench: Benchmarking Personalized Proactive Interaction in Egocentric Video Streams Authors: Dongchuan Ran, Linyu Ou, Xueheng Li, Wenwen Tong, Chenxu Guo, Hewei Guo, Kaibing Wang, Lewei Lu Published: 2026-05-08 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.07299v1 Existing Multimodal Large Language Models (MLLMs) remain primarily reactive, failing to continuously perceive environments or proactively assist users. While emerging benchmarks address proactivity, they are largely confined to alert scenarios, neglect personalized context, and fail to evaluate the precise timing of human-machine interactions (HMI).In this paper, we introduce EgoPro-Bench, a novel benchmark for training and evaluating proactive interaction capabilities based on streaming egocentric videos; it comprises 2,400 videos in the evaluation set and over 12,000 videos in the training set.Unlike previous works, EgoPro-Bench leverages simulated user profiles to generate diverse user intentions and to construct high-fidelity HMI data across 12 distinct domains.Subsequently, we propose a specialized evaluation protocol and metrics, train proactive interaction models designed for efficient reasoning and low-latency interaction on streaming video data, and conduct comprehensive evaluations.Furthermore, we introduce an interaction principle termed "short thinking, better interaction", which allocates a limited token budget prior to intent recognition, thereby enhancing interaction performance.The experiments demonstrate that EgoPro-Bench substantially enhances the intention understanding capabilities of MLLMs and enables accurate identification of appropriate timings for HMI, thereby laying a solid foundation for next-generation user-centric proactive interactive agents. --- ### Spectrum-Adaptive Generalization Bounds for Trained Deep Transformers Authors: Mana Sakai, Masaaki Imaizumi Published: 2026-05-08 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2605.07297v1 Understanding why trained Transformers generalize well is a fundamental problem in modern machine learning theory, and complexity-based generalization bounds provide a principled way to study this question. While existing norm-based bounds for Transformers remove the explicit polynomial dependence on the hidden dimension, they typically impose fixed norm constraints specified a priori and can exhibit unfavorable exponential dependence on depth. In this paper, we derive spectrum-adaptive post hoc generalization bounds for multi-layer Transformers. Under layerwise spectral norm control, the bounds are expressed in terms of layerwise Schatten quantities of the query-key, value, and feedforward weight matrices. Since the Schatten indices need not be fixed a priori and can instead be selected after training, separately for each matrix type and layer, the bounds adaptively trade off spectral complexity against the dimension- and depth-dependent factors according to the learned singular-value profiles. Empirical comparisons of BERT-adapted proxies for the leading complexity factors suggest that the proxies induced by our bounds grow more slowly with depth and hidden dimension than the corresponding norm-based proxies. Overall, our results provide a complexity-based perspective on how the spectral structure of trained Transformers is reflected in generalization analyses. --- ### On the Complexity of the Matching Problem of Regular Expressions with Backreferences Authors: Soh Kumabe, Yuya Uezato Published: 2026-05-08 Categories: cs.DS, cs.CL arXiv: https://arxiv.org/abs/2605.07289v1 ReDoS is a well-known type of algorithmic complexity attack, where an adversary supplies maliciously crafted strings to a regular expression matching engine, aiming to exhaust computational resources of systems. Even quadratic-time behavior in matching engines has been exploited in successful attacks, as exemplified by major outages at Stack Overflow (2016) and Cloudflare (2019). These incidents motivate a fundamental question: Is it possible to construct matching engines that are provably efficient, running in (near-)linear time in the length of the input string? For classical regular expressions (REGEX), Thompson's construction yields a linear-time algorithm. However, practical engines support powerful features such as backreferences, which strictly extend the expressive power of REGEX but unfortunately increase the risk of ReDoS attacks. This paper investigates the fine-grained complexity of the string matching problem for regular expressions with backreferences (REWBs). Specifically, we consider $r$-use $k$-REWBs. On the hardness side, we show that the string matching problem for $k$-REWBs cannot be solved in $O(n^{2k-ε})$ time for any $ε> 0$ under SETH. We also prove that this problem is \textbf{W(2)}-hard when parameterized by the length of the REWB expression, strengthening the previous \textbf{W(1)}-hardness. Moreover, we prove that this problem for $2$-use $2$-REWBs cannot be solved in $n^{1+o(1)}$ time unless the triangle detection problem can be solved in that time. On the algorithmic side, we present an $O(n \log^2 n)$-time algorithm for $1$-use REWBs, which significantly improves upon the recent $O(n^2)$-time algorithm by Nogami and Terauchi (MFCS, 2025). Our algorithm employs several techniques including suffix trees, transition monoids of REGEXes, factorization forest data structures, and periodicity of strings. --- ### Sword: Style-Robust World Models as Simulators via Dynamic Latent Bootstrapping for VLA Policy Post-Training Authors: Jiaxuan Gao, Yongjian Guo, Zhong Guan, Wen Huang, Wanlun Ma, Xi Xiao, Junwu Xiong, Sheng Wen Published: 2026-05-08 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.07288v1 The integration of Vision-Language-Action (VLA) models with World Models has gained increasing attention. One representative approach treats learned World Models as generative simulators, enabling policy optimization entirely within "imagination." However, when deployed as simulators for specific environments such as the LIBERO benchmark, existing World Models often suffer from poor generalization and long-horizon error accumulation. During closed-loop rollouts, these models are highly sensitive to initial-state perturbations; minor changes in color, illumination, and other visual factors can trigger cascading hallucinations, leading to severe blurriness or overexposure. Moreover, long-horizon error accumulation further degrades the quality and fidelity of predicted future states. These issues limit the reliability of World Models as simulators. To mitigate these problems, we propose Sword, a robust World Model framework. Our method introduces Structure-Guided Style Augmentation to disentangle the visual textures of interactive environments from task-relevant dynamics, thereby improving generalization. We further propose Dynamic Latent Bootstrapping, which maintains consistency between training and inference while keeping memory consumption low. Extensive experiments on the LIBERO benchmark show that our method significantly outperforms the baseline WoVR in terms of generalization, generation quality, robustness, fidelity, and the success rate of reinforcement-learning post-training for VLA models. --- ### SplatWeaver: Learning to Allocate Gaussian Primitives for Generalizable Novel View Synthesis Authors: Yecong Wan, Fan Li, Mingwen Shao, Wangmeng Zuo Published: 2026-05-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.07287v2 Generalizable novel view synthesis aims to render unseen views from uncalibrated input images without requiring per-scene optimization. Recent feed-forward approaches based on 3D Gaussian Splatting have achieved promising efficiency and rendering quality. However, most of them assign a fixed number of Gaussians to each pixel or voxel, ignoring the spatially varying complexity of real-world scenes. Such uniform allocation often wastes Gaussian primitives in smooth regions while providing insufficient capacity for fine structures, complex geometry, and high-frequency details. This motivates us to predict region-dependent primitive cardinalities rather than impose a fixed primitive budget everywhere, enabling a more expressive 3D scene representation. Therefore, we propose SplatWeaver, a generalizable novel view synthesis framework that is able to dynamically allocate Gaussian primitives over different regions in a feed-forward manner. Specifically, SplatWeaver introduces cardinality Gaussian experts and a pixel-level routing scheme, wherein each expert specializes in producing a specific number of primitives from 0 to M, and the routing scheme coordinates these experts to adaptively determine how many Gaussian primitives should be allocated to each spatial location. Moreover, SplatWeaver incorporates a high-frequency prior with attendant guidance module and routing regularization to stabilize expert selection and promote complexity-aware allocation. By leveraging high-frequency cues, the routing process is encouraged to assign more Gaussian primitives to fine structures and textured regions, while suppressing redundancy in smooth areas. Extensive experiments across diverse scenarios show that SplatWeaver consistently outperforms state-of-the-art methods, delivering more faithful novel-view renderings with fewer Gaussian primitives. Project Page: https://yecongwan.github.io/SplatWeaver/ --- ### Sparse Random-Feature Neural Networks with Krylov-Based SVD for Singularly Perturbed ODE Authors: Kevin Kurian Thomas Vaidyan, Siddharth Rout Published: 2026-05-08 Categories: math.NA, cs.LG, physics.comp-ph arXiv: https://arxiv.org/abs/2605.07286v1 Random-feature neural networks (RFNNs), including architectures with fixed hidden layers and analytically determined output weights, offer fast training but often suffer from issues due to dense representations of the hidden layer activation. Their reliance on dense feature mappings and least squares solvers can limit scalability and numerical stability, particularly for high-dimensional or stiff systems. Specifically, the activation matrix is observed to be low-rank and extremely ill-conditioned. In this work, we propose a sparse framework for RFNNs that integrates structured sparsity into the hidden layer activations that increases the rank and employs Sparse Singular Value Decomposition (sSVD) for solving the resulting linear least squares problem scalably and efficiently while catering to the bad condition number. We explore the theory behind Lanczos-Golub-Kahan Bidiagonalization technique for sparse SVD and conduct some experiments to identify some limitations and justify the requirement for orthogonalization step in our application. Then, we demonstrate that the proposed method maintains or improves solution accuracy for solving the benchmark one-dimensional steady convection-diffusion equations case having stronger advection, while achieving substantial gains in training efficiency and robustness compared to standard dense implementations. --- ### Same Targets, Different Computation: How Post-Training Divides Work Across Model Layers Authors: Yifan Zhou Published: 2026-05-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.07284v2 A late-layer change learned during post-training may work on the base model's earlier state, or it may depend on earlier computation learned with it. We distinguish these cases with a four-cell diagnostic that crosses base or descendant upstream states with base or descendant late stacks. A large late-stack effect need not imply strong upstream dependence. On math prompts, OpenMath2's late stack changes the target margin by +3.43 logits after base upstream state and +3.28 after its own, giving a near-zero interaction. Several instruction-following descendants of the same Llama-3.1-8B base show greater dependence, while controlled code and biomedical continuation-training runs sit near zero on the common support; seven released descendants span -0.54 to +2.20 logits. Because those checkpoints differ in many ways, we then isolate one training property. Two LoRA fine-tunes learn identical target responses requested either by familiar natural-language instructions or by newly learned nonce codes. Changing only this cue-to-response relation increases upstream dependence by +5.56 logits on Qwen3-4B and +4.18 on Llama-3.1-8B, with a positive paired shift in all six model-by-seed runs. The interaction is also positive in all five released base/instruction pairs we test, and late-stack replacement changes the full-vocabulary argmax in about half of events. Post-training can therefore organize the same target behavior with different dependencies between earlier and later computation. The diagnostic measures local next-token compatibility, not free-running component transplantation. ---